Unauthorized artificial intelligence tools are being used extensively across U.S. hospitals and health systems, including for direct clinical care, according to a new survey released by Wolters Kluwer Health.
The trend, known as “shadow AI,” involves the use of chatbots and apps without official institutional oversight, raising significant concerns regarding patient safety, data privacy, and regulatory compliance.
The survey of 518 healthcare professionals and administrators found that 40% of respondents have encountered unauthorized AI tools in their organizations. Nearly 20% of those surveyed admitted to using these unsanctioned tools, and one in 10 reported using them for direct patient care.
“Doctors and administrators are choosing AI tools for speed and workflow optimization, and when approved options aren’t available, they may be taking risks,” said Yaw Fellin, a senior vice president at Wolters Kluwer Health. Fellin described the issue as a governance challenge that requires leaders to “close the policy gap” and ensure only validated, secure tools are used in clinical settings.
Drivers and Risks
The primary motivation for adopting shadow AI is the need for faster workflows, cited by half of the respondents. For medical providers, curiosity, experimentation, and better functionality were also key factors.
However, the lack of formal governance has created a rift between policy and practice. The survey revealed that administrators are three times more likely (30%) to be involved in AI policy development than clinicians (9%), yet clinicians reported a higher awareness of existing policies than administrators.
Patient safety ranks as the top concern for both providers (25%) and administrators (26%), followed by fears of inaccurate outputs and data breaches.
Approximately 23% of healthcare professionals expressed specific worries about privacy and security risks, highlighting the potential for unauthorized access to sensitive medical data.
Future Outlook
Despite the risks associated with unsanctioned tools, healthcare workers remain overwhelmingly optimistic about the technology’s potential. Nearly 90% of respondents believe AI will significantly improve healthcare within the next five years.
Currently, more than half of healthcare professionals already use or rely on AI tools for their work, with data analysis being the most common application.
Scott Simeone, Chief Information Officer at Tufts Medicine, noted that while generative AI shows high potential, its successful scaling depends on organizational maturity. “Health systems need enterprise-grade controls, transparency, and literacy—so clinicians and patients understand when AI is supporting decisions,” Simeone said.
The survey was conducted online by CITE Research in December 2025.